# Cinderva AI — intelligence with a paper trail

## Product idea

Cinderva is a proposed **receipt-backed AI task network**. The useful innovation is the combination of a clear task contract, inspectable checks, portable run receipts and, later, optional Solana records. It is a product direction, not a claim that no one has explored these techniques before.

A user defines a bounded task and an accepted output format. An AI worker runs with declared tools. A verifier records whether explicit acceptance checks passed. The user keeps a receipt containing the brief, output, evidence, evaluation version and execution metadata. Optional worker signatures identify a registered worker; optional on-chain digests make a record easier to reference across services.

Example: a small business asks for a three-bullet product summary. The packet names the allowed sources and limits the response length. Structure checks can validate the output format; separate citation checks and human review assess whether the summary is supported. Payment release in a future paid-job system would depend on the agreed rules and dispute process, not on a vague “AI approved” label.

## What the prototype actually does

The browser receipt lab accepts editable sample JSON. It checks a title, three bounded bullet strings and an attached source ID. It records those format results, hashes a defined JSON encoding with SHA-256, exports JSON and detects edits to a receipt’s stored contents. An import action checks an exported file against its recorded digest.

It runs locally, without an AI provider, wallet connection, transaction, account, on-chain anchor or payment. The attached source identifier is not a verified citation. An unsigned receipt can be recreated by anyone, so a matching digest is **not** proof of origin, model execution, accuracy, honesty or useful work. These limits are visible in the UI.

## Proposed product stages

1. **Useful execution:** a server-side model adapter, bounded tool permissions, job queues, customer data controls and explicit per-task pricing.
2. **Better evidence:** versioned evaluations, citation-resolution checks, registered worker keys and signed receipts. Human review for subjective or higher-stakes tasks.
3. **Portable records:** optional Solana digest anchoring. Keep private inputs and outputs off chain; a public hash can still leak information about predictable content and needs a privacy review.
4. **Paid tasks:** capped prepaid jobs or escrow, with acceptance rules, disputes, refunds and operational controls built and tested first.
5. **Optional CNDV integration:** a defined task-access/credit mechanism, with a sustainable service-cost model and published customer terms. Coin creation itself supplies none of these services.

## Why a token might fit

The proposed token role is optional prepaid task access and a shared participation layer around the network. Its usefulness would depend on actual services and clear pricing. CNDV would not be an equity interest, guaranteed-return instrument or automatic claim on project revenue. Staking yields, automatic buybacks and paid-validator rewards are not part of this prototype or approved economics.

Launching a tradable coin before services exist must not be presented as a working decentralised AI network. The website therefore identifies the live local demo and every planned product stage separately.

## Founder ownership

Your requested target is **5% of actual total supply**, with no additional hidden founder/team treasury allocation proposed here. The public curve/liquidity path is not the same as community ownership at genesis. Acquisition, actual addresses and any lock policy must be finalised and disclosed. See `FOUNDER_POLICY.md` and `LAUNCH_GUIDE.md`.

## Brand

Name: **Cinderva AI**. Ticker: **CNDV**. The visual system reuses the selected Bloom layout, type scale, sculptural flower and responsive patterns, recolored in blood red, oxblood and pale rose. The name and ticker need final brand/domain review before irreversible metadata is created.